a16z Releases Latest Consumer AI 100: Beyond Traffic, Who Is Really Paying for AI?
Personal card spending data added for the first time: Who spends $903 a month, where does the money flow, and how can AI get everyday people to keep using it?
Consumer AI is booming, but the bulk of the money is still coming from people using it to write code, produce content, and get work done.
a16z's newly released seventh edition of the top 100 consumer AI apps, authored by Olivia Moore, incorporates US personal card spending data tracked by YipitData for the first time. The findings are stark: the top 1% of spenders spend an average of $903 per month on AI tools, while the median spender spends just $25 per month.
The next competition is getting everyday people to use AI daily. Relying solely on "boosting productivity" is not enough: products must offer experiences people actually want and return to, while finding monetization models that broader audiences can afford.
Who Is Really Paying? Spending Concentrates Among a Minority of Work and Creative Users
Paying Users Are Growing, but Spending Remains Highly Concentrated
In the US survey cited in the report, nearly half of surveyed consumers reported having used AI, with approximately 25% of overall consumers using it daily. In YipitData's sample of qualified US personal electronic receipts, the share of consumers subscribing to at least one of ChatGPT, Claude, or Gemini was approximately 4.5%, up from 2.1% a year earlier. The usage survey and subscription data stem from different samples; while the paying base is expanding, personal card spending reveals that growth remains deeply uneven.
At the same time, spending among heavy spenders is accelerating. In the observed consumer card spending sample, consumer AI exhibits a pronounced power-law distribution (referring here to spending being heavily concentrated among a small fraction of paying users):
- Extreme concentration: The top 10% of paying users contribute roughly half of the total spending in the observed sample; the top 1% alone accounts for 19.5% of total spend—more than the combined total of the bottom 50% of paying users (16.6%).
- A wide spending chasm: The top 1% of spenders spend an average of $903 per month, up about 80% over the past year and a half; average monthly spending per user in the top 10% tier grew 31%, and in the top 40% tier grew 10%. Incremental spending is distinctly more concentrated among the heaviest users. Meanwhile, the median across all paying users remains around $25 per month, showing little long-term change.
- What are they actually buying? This group's purchasing propensity tilts heavily toward tools used for building, creating, and completing work, rather than casual chat. This aligns closely with prosumer (individuals purchasing professional tools under personal accounts for work or creation) behavior. Statistics show that the top 1% spending tier exhibits a much higher propensity to buy automated workflow platforms (e.g., n8n), AI developer infrastructure (e.g., fal, Nous Research), and high-barrier creative tools (such as video generator Higgsfield, design platform Figma, avatar generator HeyGen, and intelligent workflow tool Manus). Here, "higher purchasing propensity" refers to the ratio of a tool's purchase rate within the top 1% spending tier relative to its purchase rate across all AI buyers, and does not mean that an absolute majority is buying it. For example, for the workflow tool n8n, the purchase rate among the top 1% of spenders was 22.9 times that of all AI buyers, yet its absolute purchase rate within the top 1% was only 1.43% (compared to 0.06% across all buyers).
In the companion interview, Olivia noted a common adoption path: individuals start using work-related AI tools on their own, find them valuable, and bring them into their teams, after which tool vendors follow up with privacy, security, and team plans. This is product-led growth (PLG, Product-Led Growth, where users adopt a tool organically through product experience and introduce it to colleagues or teams). This indicates that spending categorized as "personal consumer" is not equivalent to pure entertainment consumption; many work-adjacent use cases are initiated out-of-pocket by individuals. Enterprise-level unified software procurement is not included in this personal card dataset.
Under this highly concentrated spending structure, paying for multiple tools is also uncommon. YipitData's sample shows that among all users already paying for at least one AI product, only 13% pay for at least one additional AI tool. For most people in the sample, spending involves only a single AI tool. New tools must clearly articulate what additional problem they solve before people trying them will be willing to pay an extra subscription.
Where Traffic Rankings Don't Look, People Are Also Paying
To clarify what different data dimensions indicate, this report actually consists of three distinct, standalone Top 50 lists:
| List Dimension | Metric & Data Source | Methodology Notes |
|---|---|---|
| Web Top 50 | Monthly website visits (Similarweb) | Measures desktop and web traffic popularity; does not include in-app usage |
| Mobile Top 50 | Monthly active users (MAU) (Sensor Tower) | Measures mobile app active scale; cannot be simply added to web visits |
| Personal Spending Top 50 | Observed US personal card spending (YipitData) | Tracks individual payments and high-ticket demand; does not represent companies' total global revenue |
These three lists track web visits, mobile activity, and personal card payments separately; they are not a merged ranking of 100 deduplicated products, nor are they a model capability leaderboard evaluating underlying technical benchmarks. Both AI-native products and existing legacy products qualify for inclusion, provided generative AI forms a core part of the user experience.
A total of 11 products appeared on this list series for the first time in this edition, a series low. Products primarily dedicated to adult content were removed from the general rankings; the report estimates that had they been included, they would account for over 20% of the top 50 web spots.
Different access points and paying demographics yield diverging ranks between website popularity and individual payments: among the top 50 AI companies by personal card spending, as many as 29 did not make the top 50 on either the web or mobile traffic lists. This highlights that relying solely on web and mobile visits overlooks certain product form factors (such as desktop applications or tools embedded in messaging apps); looking only at browsers and mobile apps misses the usage and payments occurring within these channels.
In the personal card spending rankings, OpenAI ranks No. 1, Anthropic ranks No. 2, Canva ranks No. 3, and Superhuman ranks No. 4. Notably, Gemini does not appear in the top 50 of the spending list because its consumer revenue cannot currently be reliably separated from Google's broader consumer operations, not because its ranking is low or unprofitable.
Furthermore, because the personal card spending list (tracking US personal cards) and the traffic lists measure different population samples and observation windows, rank discrepancies cannot be used to directly infer conversion rates, nor do they prove higher revenue per user. However, this comparison effectively reveals where traffic metrics may miss high-spending demand:
| Product | Personal Card Spending Rank | Web Traffic Rank |
|---|---|---|
| Higgsfield | No. 5 | No. 28 |
| Replit | No. 6 | No. 48 |
| Suno | No. 7 | No. 19 |
To evaluate whether a tool commands paying demand, looking beyond web traffic to see what work workflows it enters and whether users are actually paying is essential.
Where is it popular, and where is it paid for?
ChatGPT
ChatGPT appears across all three lists.
Data is drawn from this edition's three overall lists. The spending list tracks merchants, with OpenAI and Anthropic corresponding to ChatGPT and Claude, respectively; tracking periods and populations differ across lists, so they cannot be used to calculate paid conversion rates.
Across all tracked products, only 7 companies appear simultaneously across web traffic, mobile monthly active users, and personal card spending: beyond ChatGPT, which holds the No. 1 spot across all three lists, the group includes model provider Claude (Spending No. 2, Web No. 3, Mobile No. 12), music generation platform Suno (Spending No. 7, Web No. 19, Mobile No. 47), two native applications—Perplexity (Spending No. 9, Web No. 12, Mobile No. 25) and Photoroom (Spending No. 38, Web No. 36, Mobile No. 45)—along with two established productivity tools: Canva (Spending No. 3, Web No. 4, Mobile No. 4) and Notion (Spending No. 17, Web No. 7, Mobile No. 28).
The Three Major Assistants: ChatGPT Serves Broad Audiences, Claude Focuses on Professionals, Gemini Relies on Its Ecosystem
As the foundational layer of the generative AI industry, the contest among the flagship assistants from OpenAI, Anthropic, and Google has drawn the most attention. What is compared here is the consumer-facing product positioning and subscription tiers of the three assistants. Latest data indicates that the three are diverging into distinct commercial and product models.
ChatGPT: Scale Leader, Serving Both Mass and Work Scenarios
Whether in web traffic (roughly 2x runner-up Gemini and 6x Claude) or mobile monthly active users (MAU, active users who open the app at least once a month), ChatGPT maintains a commanding lead. On mobile in particular, ChatGPT's MAUs reach 2.5x Gemini's and 14x Claude's, contrasting sharply with the narrower web ratio. In YipitData's tracked US personal subscription sample, ChatGPT's paid subscriber base is roughly 3x that of each of the other two.
In observed OpenAI merchant spending, women contribute approximately 41.5% and men about 58.5%, spanning a broad range of household income brackets. Across US states, with the exceptions of California, Massachusetts, Montana, and Washington, D.C., where Anthropic leads, ChatGPT leads in observed spending across the other 47 states. A more balanced spending distribution does not imply user headcounts are distributed in identical proportions. In terms of product iterations, OpenAI's dozens of recent updates span daily life features like image generation and personal financial assistants, as well as team-focused workspaces like ChatGPT Work.
Additionally, in the tracked US personal subscription sample, only 8% of ChatGPT subscribers also subscribe to Claude. That is, the vast majority of ChatGPT subscribers in the sample do not co-subscribe to Claude. This proportion is observed from the ChatGPT subscriber side and cannot be inverted to represent co-subscription rates among Claude users; nor do both products need to fight solely over the exact same use cases for the same audience.
Claude: Professional Demand Drives Higher Pricing Tiers
Compared to the first edition of the list three years ago, Anthropic's Claude has demonstrated powerful momentum, ranking in the top three for web traffic and at one point surpassing Gemini in personal paid subscribers. This wave of growth accompanied new product rollouts (such as Cowork for document collaboration and design tool Claude Design), new models (such as Fable 5), and public attention surrounding the Pentagon controversy. It also briefly topped the US free mobile app charts. After Google restructured its cloud storage plans, Gemini's subscription curve ticked upward, and the two have drawn close in the latest observation period: active personal subscription shares in the qualified US sample stood at 1.09% for Gemini and 1.07% for Claude.
Anthropic has explicitly stated that it does not serve ads within Claude, monetizing instead via subscriptions:
- Among its active individual subscribers, a notable 7.3% subscribe to the highest-tier Max plan starting at $100 per month;
- In contrast, users on $100-plus top-tier plans represent roughly 1% for both ChatGPT and Google.
In observed spending breakdowns, Claude shows a distinct skew toward higher-income and professional tool purchasers: men account for more than three-quarters of spending, and households earning over $150,000 contribute a high share of spending; over 60% of product announcements in the past six months focused on code review and professional work contexts. However, this focus on professional users has recently shown divergence: desktop monitoring panels indicate that its daily average sessions dipped over the last two observed months, while electronic receipt samples show that gross new customer additions slowed and churn edged higher. Meanwhile, ChatGPT re-accelerated following the release of a new model family and workspace products.
Gemini: Legacy Ecosystem Yields Subscriptions, Multimodal Expands Daily Uses
Google possesses extensive distribution channels. Google previously migrated and upgraded its legacy paid cloud storage plans into premium plans bundling AI capabilities, and this migration of existing accounts drove an uptick in measured paid penetration. This increase includes legacy paying users transitioning over and cannot be interpreted entirely as newly generated AI willingness to pay. In recent product announcements, nearly half of Google's new features centered on consumer and mass creative domains (such as music models and multimodal generation), leveraging Search and system ecosystems to continually expand touchpoints.
Personal Agents: From Answering Questions to Booking and Shopping
Moving the Front Door to Messaging Apps to Make Delegation Easier
In the sixth edition, early agents (AI assistants capable of planning steps and calling tools to execute tasks on behalf of users) like OpenClaw were seen as having a steep learning curve for most everyday users. Half a year later, startup assistants such as Instinct, Tomo, Poke, Lindy, and Town have reported hundreds of thousands of users; tech giants have also entered the fray, launching Meta Muse, OpenAI Dots, and xAI Grok Bot. The companion interview notes that some newer assistants were not yet included in this edition's main rankings, and the report tracks their early progress separately.
These assistants aim to function like human executive assistants: you delegate tasks by sending a text, emailing, or making a call. iMessage has become a popular entry point, where Instinct got its start, and DoorDash launched an SMS food-ordering assistant. Users can start delegating without having to adapt to a new website first. Assistants targeted at professional work or specialized tasks often retain dedicated web and mobile apps for completing deeper, complex workflows.
What are these assistants actually helping people do? According to an analysis of seven early user community group chats by evaluation outfit AssistantBenchmark (within a sample of 1,502 messages, 401 mentioned at least one task), coding and technical automation remained the most frequently discussed domain (103 mentions), followed closely by travel arrangements (66 mentions), everyday shopping (58 mentions), and personal finance (56 mentions). This reflects what early users talked about in group chats, including questions and ideas; it records discussion volume and cannot be taken as counts of completed tasks.
This shift points toward an intriguing commercial avenue for consumer AI: moving from monthly software subscription fees to experimenting with transaction-based takes on delegated execution.
- For example, SMS assistant Instinct, which currently remains invite-only, saw founder Noah Shinn state that it attracted over 100,000 users in three weeks. According to team disclosures, 40% of users bind personal credit cards within three weeks; once purchasing behavior begins, transacting users spend an average of approximately $1,300 per month through Instinct. The founder reported that over $1 billion in annualized transaction volume is flowing through Instinct, with roughly half attributed to travel. This $1,300 represents purchases of goods and services such as flights and hotels; the $1 billion annualized transaction volume is not Instinct's revenue.
- Meta Muse reached 250,000 daily active users in its first week and logged over 5 million cumulative downloads in less than a month, currently available only in the US and Canada. In their respective first 22 days in these markets, Threads logged 15.71 million downloads compared to 5.38 million for Muse. The report frames the core challenge for new consumer products as overcoming user indifference: once the novelty fades, users must have a compelling reason to keep opening it. Download figures illustrate initial interest in trying a product; sustained usage depends on retention.
Connected, Executable, and Safe Enough to Trust with Delegation
As products expand into everyday routines, deploying autonomous agents faces several interconnected practical challenges:
- Platform rules and API partnerships: As agents begin making purchases online on behalf of users, platforms are paying significantly closer attention to transaction attribution and funnel ownership. For instance, Amazon cut off Meta's assistant from automated access to its store within two weeks; meanwhile, Shopify, food delivery, and travel booking platforms have opted to partner via official APIs, and companies like DoorDash launched their own SMS food-ordering agents. Platform partnerships address the technical question of whether integrations and transaction flows can connect successfully.
- Trust mechanisms and delegation thresholds: Connecting to personal email, work files, and credit cards provides agents with deeper contextual understanding and execution capability, but it also makes users far more sensitive to whether the agent might execute unconfirmed actions, misroute sensitive information externally, or breach confidentiality boundaries across multiple contexts. Interview guests pointed out that people instinctively know what private details can be shared and what must be kept confidential in daily social interactions; whether software can reliably manage these boundaries dictates what users are willing to entrust to it. Thus, casually testing a tool does not mean users are ready to entrust it with critical matters over the long term; building lasting daily habits hinges on predictable trust mechanisms.
- Differences between single-player tools and network effects: Integrating external merchant and service APIs fundamentally allows an individual user to accomplish more tasks with an agent, representing a lateral expansion of single-player tool capabilities. Yet across the history of the consumer internet, major businesses were built on interpersonal networks or transactional marketplaces. Social products like Threads must achieve a critical density of interaction in very short order for the network itself to become compelling; personal assistants, by contrast, deliver utility even with just a single user. Precisely because personal assistants house highly sensitive individual context, that context cannot be casually ported into multi-party interactions. Connecting more tools and building interpersonal networks are two distinct challenges that must be solved separately.
- Early days for monetization: While both Instinct and Muse plan to monetize via transaction cuts or intermediary fees in the future, neither had formally implemented this at the time of publication; similarly, checkout flows initiated inside ChatGPT do not currently charge transaction fees to users or merchants.
To discern where value truly originates across different product formats, current consumer AI exploration scenarios can be compared against their underlying mechanisms:
Connects to email, calendar, and merchants so one person can get more done; friends don't need to join first.
The product only becomes compelling when enough people actively communicate, share, or match; more participants can bring more connections and content.
Connects supply and demand and must fulfill transactions; an assistant understanding user needs better is only one step.
In two-sided matching domains like dating and recruiting, or marketplaces like e-commerce and real estate, interviewees noted that AI can help people express needs more clearly and discover suitable counterparts. However, products still require sufficient buyers and sellers or matching candidates, as well as the ability to fulfill transactions; an assistant's delegation capability does not automatically create these networks.
Startups: Beyond the Foundation Model, What Keeps Users Around?
Mature software products have established positions on the leaderboards: Canva and Notion rank in the top ten on web, Figma ranks in the top fifteen, and Superhuman (formerly Grammarly) sits in the top half of the web list and 4th on personal spending. Google placed five web products on the ranking: Gemini, NotebookLM, AI Studio, Labs, and Antigravity.
These incumbent products present two possibilities: they may introduce broader audiences to AI first, prompting users with deeper demands to graduate to native AI tools; or they may fulfill those needs directly, becoming formidable long-term competitors to startups. From startups that continue to win users, the report identifies four competitive strategies:
Specialized Models and Workflows for Concrete Creative Needs
Competitive dynamics vary across audio, image, and video; survival is not as simple as "possessing a proprietary model":
- Audio and music: Olivia observed that foundation model providers must divide resources between coding and general intelligence while navigating music copyright hurdles; these trade-offs leave space for dedicated tools. Music generation platform Suno (Web No. 19, Spending No. 7) and voice synthesis product ElevenLabs (Web No. 25, Spending No. 10) both secured spots across web traffic and personal card spending lists, demonstrating enduring demand for specialized creation.
- Image generation: While general foundation models (such as ChatGPT's native image generation and Google's image tools) absorb large volumes of casual generation traffic, the interview noted that professional creators still require refined aesthetic control, granular adjustments, and end-to-end editing workflows. For example, Midjourney did not make the top 50 on either web or mobile traffic in this edition, yet ranked 11th on personal card spending. This contrast underscores that mass web visits do not capture the full scope of paying demand. Guests explained that professional users care not just about generation, but about visual aesthetics, how details can be tuned, and what assets can be saved and iteratively edited throughout the creative process.
- Video generation: Video entails multiple complex variables including visual quality, sound, audio-video synchronization, and stylistic coherence. Tools like KlingAI (Spending No. 26), HeyGen (Spending No. 23), and Topaz Labs (Spending No. 27) have all carved out positions on the spending list.
Additionally, guests noted that as more powerful foundation models learn to orchestrate external creative tools via tool calling, whether vertical tools are easily discoverable, identifiable, and callable by various agents is emerging as a new lens for evaluating their future distribution channels.
Cross-Model Orchestration to Pick the Right Model for the Task
Foundation model labs inherently aim to retain users and consume their own compute tokens (the fundamental data units large language models process); in contrast, Cursor, which supports calling models from multiple labs, and OpenRouter, which routes multi-model access for developers, compose models around user task requirements, establishing a product-level competitive edge. In coding and productivity, tools including Lovable (Web No. 14), Cursor (rising from No. 41 to No. 35), Base44 (debuting at No. 46), and Replit (Web No. 48) made the web traffic list.
Serving Specific Audiences to Address Needs Overlooked by General Assistants
OpenEvidence, a clinical decision-support tool built specifically for physicians, climbed to No. 47 on the web list, while privacy-focused Venice debuted at No. 43. For these professional or niche users, source reliability, privacy protection, and alignment with specialized workflows matter just as much as underlying model benchmarks.
Content policies also partition the market. Use cases that mainstream assistants decline to serve still have genuine demand, opening opportunities for niche tools to serve those audiences. Moving adult-oriented products out of the general ranking in this edition reflects a classification shift; the underlying demand did not disappear.
Rebuilding the Interaction to Accumulate the User's Own Work Context
Interaction opportunities extend well beyond the chat box. Legacy interfaces across Google Docs, Calendar, and Gmail leave room to rethink how work gets done; while incumbents must protect existing habits and revenue lines, startups can redesign experiences with greater agility. Plaud packaged meeting recording and transcription into portable hardware, reaching No. 16 on the spending list through a "hardware + AI membership" model.
Hardware is not a permanent safe harbor either. The report notes that Meta announced plans to expand Muse into smart glasses and AI digital pets, and OpenAI's acquisition of io Products indicates hardware ambitions. These moves show that tech giants are similarly pursuing access points beyond the chat interface.
Software also offers cumulative value: gradually understanding user context around their workflows. As noted in the discussion (see companion interview at 39:29), foundational model upgrades are a double-edged sword for startups: if a product is merely a thin wrapper around a general model, model improvements can easily subsume its functionality; but if a product has integrated deeply into a user's workflow, established its own business processes, and accumulated distinct personal context, stronger base models enable that product to accomplish even more.
In the interview, Olivia cited her own experience using Town to draft emails (43:22): the assistant connects to Slack, email, and Google Drive, absorbs personal and organizational context, and creates a playbook (a guide capturing personal habits and communication style) detailing who she is and how she communicates. She found the drafts sounded more like her own voice, reducing time spent editing emails. This illustrates her personal impression and highlights how applications can minimize rework by compounding context over time. Switching to another model does not automatically transfer these long-accumulated personal expressions and working background to another application.
Business Models: Beyond Subscriptions, How to Serve More Everyday People?
Reflecting on the mobile internet boom, the core business playbook was "users enjoy free services while platforms monetize via advertising or transaction facilitation." In the revenue breakdown of traditional tech giants, advertising accounts for over 97% of Meta's revenue and more than 70% of Alphabet's; Amazon's Prime subscription service, despite its immense scale, represents less than 10% of total company revenue.
Today's consumer AI, however, displays a vastly different monetization landscape:
- All 44 AI-native web products on the ranking offer at least one monetization model. 84% offer personal subscriptions, 64% offer usage-based or add-on credit billing, 14% feature advertising, and 2% incorporate transaction take rates. Companies can deploy multiple models simultaneously; these figures measure model adoption rates, not revenue shares.
- Advanced features and higher usage limits frequently require a subscription or extra credits. For users with clear work deliverables, this expense maps straightforwardly to output; for casual users who have yet to build habitual usage, being asked to commit to payment upfront creates a much higher barrier to entry.
Covering Inference Costs First: Subscriptions Still Fit Well-Defined Needs
A critical reality is that model inference incurs real compute costs. Traditional consumer software could expand its user base for free before layering on ads or commerce; AI consumes compute with every interaction, whereas advertising and transaction cuts often first require scale. Before reaching that scale, subscriptions or credits are a natural way to cover service costs.
Across the 20 large consumer subscription products benchmarked by a16z, 65% belong to media categories like Netflix and Spotify. Users pay recurring fees for exclusive content or ad-free experiences, and AI products must similarly prove long-term value. ChatGPT entered this benchmark roughly three and a half years after launch, showing that subscriptions can support large-scale products; however, because reporting dates and subscriber definitions vary across companies, this compilation is not a synchronized, precise ranking.
Lowering serving costs is one avenue toward expanding accessible usage:
As open-source models proliferate and product architectures explore routing complex tasks to lower-cost lightweight models, marginal serving costs could theoretically improve, leaving room for free or lower-friction tiers.
Conversations Clarify Purchase Intent and May Enable More Relevant Advertising
OpenAI stated that, on a base of 1.2 billion weekly active users, its advertising business reached a $1 billion annualized revenue run rate. An annualized run rate extrapolates the current pace of revenue and does not represent actual cash collected over the preceding twelve months. Separately, media outlets including NBC have reported that roughly 50%–60% of licensed US physicians use OpenEvidence, which currently monetizes partly through ads.
In the interview, guests explored the potential of marrying conversational formats with advertising: compared to single keyword queries in traditional search engines, users engaging in multi-turn dialogues with AI progressively clarify and articulate highly specific purchasing intents. This theoretically allows recommended products or solutions to achieve much higher relevance. For instance, Josh Elman shared a personal shopping experience (see companion interview at 27:36): he was searching for a specialized keychain that easily detaches a valet key, which proved hard to pinpoint via traditional search; after back-and-forth descriptions of mechanical details with ChatGPT, followed by adding the constraint "preferably made in the USA," he located a product made by a factory in Detroit. No ads appeared in that interaction, but it demonstrated how multi-turn conversation helps users refine high-certainty purchase intent.
Guests hypothesized that this deep intent understanding could eventually unlock higher conversion rates and premium ad pricing; yet they uniformly cautioned that introducing commercial promotions into deeply intimate daily dialogues demands strict restraint. Heavy-handed weight-loss product recommendations inside a private health conversation would quickly lead users to question whether they are receiving impartial advice or a sales pitch.
Transaction Takes Offer Opportunity, but Require Scale and Attribution
As agents progress from conversational Q&A to helping book tickets and shop for goods, defraying serving costs via transaction takes has emerged as a promising concept. Yet this model demands enormous transaction volume and reliable attribution systems; whether it can broaden everyday adoption remains to be proven by these products.
The Mass Market Still Needs Experiences Worth Spending Time On
The category matrix distinguishes between "represented," "partially covered," and "no ranked representative." OpenEvidence in health and Meituan in local services are labeled Partial, indicating they cover only narrow use cases or specific geographies rather than the broader category.
Streaming, shopping, social, travel, finance, recruiting, real estate, gaming, and dating are designated None, indicating they lack dedicated representatives in this edition's two traffic top 50 lists, not that AI products do not exist in the wild. Streaming in the chart excludes creative tools, and social excludes AI companion products, while Character.ai remains on the main list. Expanding into these categories requires tackling specific structural challenges: dating and hiring need two-sided matching, shopping and real estate require connecting supply and demand, and entertainment and social require motivating people to create, watch, and engage. A more capable conversational assistant will not automatically build these networks.
Interview guests situated the key challenge for the next wave of expansion squarely in user motivation: professional developers and creative professionals come with concrete deliverables, where AI directly saves time and amplifies output; for mass first-time users, facing a generic blank prompt often leaves them unsure what everyday problem it solves. In the real lives of everyday people, consumption is not solely about "saving time"—it encompasses entertainment, self-expression, social connection, and the tactile joy of bringing an idea from one's imagination into reality. More capable models and lower prices can lower the friction of adoption; products must still help people see what they actually want to do with them today.
Members can copy or download the full Markdown, and let their AI subscribe and ask.
Found this useful? Leave your email — every new explainer, the moment it's out. Unsubscribe anytime.